AI startups: Sell work, not software
The article argues that AI startups should sell completed work or outcomes rather than traditional software licenses, because AI can automate entire tasks. Pitching "work output" aligns better with customer value than selling tools, and AI companies should price based on results delivered instead of per-seat subscriptions.
Background
- Sarah Tavel is a well-known venture capitalist (formerly at Benchmark, now at a16z) who writes about product strategy and startup patterns. Her "sell work, not software" thesis argues that AI startups should charge for completed tasks or outcomes (e.g., "I will generate this report") rather than selling traditional software licenses (e.g., "I sell you a tool that can be used to generate reports").
- The core insight: traditional SaaS sells tools that require human labor to operate. AI can now replace that labor, so the pricing model should shift from per-seat/usage to per-outcome. A customer pays for the finished work itself.
- This matters because it reframes how investors and founders think about AI business models. Most AI startups today still use SaaS-style pricing; Tavel suggests that's a mistake — the most valuable AI companies will behave more like services businesses with software margins.
- Prior context: Tavel previously wrote about the "hamburger menu" pattern where incumbents get disrupted, and has long argued that AI compresses the value chain — eliminating the middle layer of human labor between the software and the output.